Proverbs and advice such as “don’t judge a man until you’ve walked in his shoes” and “take some time to get to know her” suggest that people’s abilities to perceive others change with increased acquaintance. Building from such beliefs, people reasonably expect that those who are well acquainted will be more likely to agree on their perceptions of one another as compared to others who do not know each other as well. That is, a group of people who are well acquainted will agree more with one another’s judgments about each of the group members as compared to a group of less well acquainted individuals. Despite the reasonableness of this belief, researchers have often failed to show an effect of acquaintance on how much people’s perceptions of others agree (Kenny, Albright, Malloy, & Kashy, 1994; Park, Kraus, & Ryan, 1997). This rift between popular belief and empirical findings can lead to one of two conclusions: either members of the general public do not understand their own perceptions of one another, or previous investigations of this phenomenon have not fully captured the richness of how person perceptions change over time. The current study was based on the latter conclusion and was designed to improve upon past studies examining the role of acquaintance on agreement of personality judgments by extending the period of time assessed to include initial acquaintance.

Empirical investigations into the role of acquaintance on levels of agreement, or consensus, in judgments of personality traits have yielded inconsistent conclusions. A review of 20 cross-sectional studies (Kenny et al., 1994), including both studies of highly acquainted individuals and nonacquainted individuals, suggested that acquaintance may lead to increased consensus with regard to the Big Five (Norman, 1963) personality traits. Of the studies reviewed, those in which consensus at zero acquaintance (no prior interaction between individuals) was examined demonstrated moderate levels of consensus for judgments of extroversion and conscientiousness and lesser levels of consensus for agreeableness, emotional stability, and culture. Consensus levels across the Big Five traits were consistently higher in studies of people at low acquaintance as compared to those at zero acquaintance. Further, agreement was higher still in studies of highly acquainted individuals than those shown in zero or low acquaintance relationships.

Conversely, the 12 longitudinal studies of consensus reviewed by Kenny et al. (1994) indicated that interjudge agreement does not increase as a function of acquaintance. In these studies, which included both experimentally constructed and naturally occurring groups, only slight increases in consensus for conscientiousness, emotional stability, and culture were shown. There were no increases in agreement for ratings of extroversion or agreeableness over time. In addition, more recent longitudinal studies examining this question over the course of seven weeks (Paulhus & Reynolds, 1995) and eight months (Park et al., 1997) also failed to show an increase in consensus as a function of acquaintance.

The inconsistency in these findings may be related to methodological differences in the ways researchers have approached this question. In order to assess agreement in person perception, reliable measurements of inter-judge agreement first must be defined. Whether defined by consensus in the judgments of others or by the correlation between one’s own judgment and the judgments of others, measures of person perception represent complex arrays of social processes. Because of this complexity, many early studies of person perception have been criticized for lacking the statistical sophistication necessary to capture the complexity of person perception (for further discussion of these arguments see Cronbach, 1955; Kenny, 1994). In an attempt to better measure person perception, the Social Relations Model (SRM; Kenny & LaVoie, 1984; Warner, Kenny, & Stoto, 1979) was developed to statistically control for biases in social judgments. The SRM allows variance in perceptions to be componentially analyzed to attribute any variance within perceptions to the perceiver, the target, or the unique interaction between those two individuals. Yet despite using this statistical procedure researchers have reached different conclusions about the function of acquaintance in agreement due to differing definitions of acquaintance.

In addition to using statistically valid measures of agreement, appropriate measures of acquaintance must be used in order to address the relationship between these constructs. When considering the effects of a dynamic construct, such as acquaintance, longitudinal or repeated measures designs are better suited to examining changes in individuals’ perceptions than are studies that compare the perceptions of various people representing different levels of acquaintance. Evidence of a relationship between agreement and acquaintance based on cross-sectional studies is inherently flawed; nevertheless, it is interesting that data examined across these individual studies have suggested an effect of acquaintance (e.g., Kenny et al., 1994) while the longitudinal studies conducted thus far have failed to show this effect (e.g., Albright, as cited in Kenny et al., 1994; Park et al., 1997).

An obstacle to examining the effect of acquaintance on agreement is reliably assessing levels of acquaintance. In person perception research, relationships have typically been classified as representing either zero acquaintance, short-term acquaintance, or long-term acquaintance. Such classifications have relied on the number and length of interaction periods individuals have had despite the fact that there may be more variation experiences within each classification than there is between them. For instance, for the purpose of a review of the literature (Kenny et al., 1994), a study of Ham radio operators and a study of college roommates were both used to represent long-term acquaintance despite the qualitative differences between the relationships of the two groups. The differences among studies that are unaccounted for within this simple classification system may lead to differences in agreement level regardless of the frequency of interactions. Differences in the situational factors that define the interactions of people, such as characteristics of friendship-based unstructured time versus task-oriented classroom discussions, have been shown to differentially affect agreement levels (Paulhus & Reynolds, 1995). Thus, the true effect of acquaintance may not have been reliably assessed when conclusions are drawn averaging across varied situations.

Improving upon classifications based strictly on time measures, researchers have operationalized acquaintance as self-reported assessment by individuals of their relationships (e.g., Kenny & Kashy, 1994; Park et al., 1997). Although self-reported acquaintance scores increased over time, these studies failed to show corresponding increased consensus. However, it should be noted that subjective ratings of acquaintance are not independent of the specific time periods in which individuals make their ratings. Willett (1989) argued that the reliability of data collected in repeated measures designs is related both to how many observations are made as well as to the interval of time between the measures. Similarly, conclusions about a particular phenomenon have been shown to be affected by the arbitrary selection of an initial timepoint as well as by the chosen duration of the observation period (Rogosa, 1988). Thus, if ratings of the subjective feeling of being acquainted are to provide an adequate measure of acquaintance, they must be gathered after appropriate passages of time that allow for different levels of acquaintance – for instance, either short-term or long-term acquaintance – to be experienced. It is possible that changes in acquaintance were detected because individuals were repeatedly asked about acquaintance levels but not provided with opportunity for agreement to increase. Moreover, in order to relate ratings of acquaintance to measures of agreement, the time points selected to collect measures of each must be appropriate for identifying change in both variables. The selected timepoints for data collection may have been adequate to detect changes in acquaintance but inappropriate for identifying changes in consensus.

In light of these shortcomings, an examination of the methodologies of two of the more compelling studies to date highlights the reason that researchers may have failed to find an effect of acquaintance on agreement expected by popular belief. In an effort to capture the complexity of the person-perception process in an unbiased way, both Albright (1990, as cited in Kenny et al., 1994) and Park et al. (1997) collected measures of agreement with SRM designs. Furthermore, in accordance with critiques of the field (Kenny, Horner, Kashy, & Chu, 1992; Kenny, 1994; Malloy & Albright, 1990) their data were collected from individuals who lived together over a period of time and interacted in naturalistic environments unconstrained by experimental protocol. Albright measured consensus levels of trait perceptions three times over the course of an academic year. However, the initial ratings were collected after five weeks of acquaintance. Similarly, Park et al. (1997) collected ratings of the acquaintance of college roommates and perceptions of personality traits four times over eight months, but the first data collection occurred after a two-week period. These lapses between actual initial acquaintance and the initial measurement most certainly allowed individuals time to become acquainted prior to the start of the studies. Although in these studies the relationship between quality measures of interjudge consensus and acquaintance were used, this relationship was not assessed during the potentially critical period of initial acquaintance, when perceivers were first forming impressions of others.

The present study was designed and conducted to examine more fully the role that acquaintance has on agreement of personality judgments during this unexplored initial period. Building from the strengths of previous investigations, it was expected that this study would help provide a clearer picture of the relationship between acquaintance and agreement. Acquaintance was not only operationalized as the passage of time over a six-month period, but also by subjective reports provided by participants at each of the data collections. Moreover, to capture as much of the acquaintance process as possible, initial measures of agreement were collected at the first opportunity (within the first week of members of the group meeting) so as to more closely approximate a zero-acquaintance level than in previous reports. In this study, the relationships between group members from their (close to) initial meeting over a six-month period were examined in a methodologically appropriate way. It was expected that this study would provide a more conclusive definition of the effect of acquaintance on agreement in personality judgments than in previous research and would bridge the gap between empirical findings and common conceptions of the acquaintance process.

Upon arrival at campus, students were compelled to take part in a time-intensive orientation program during which they participated in group discussions and activities. For this program, incoming students were randomly assigned to activity groups that were all housed in one campus dormitory. The structure of this program minimized the opportunity for informal social interactions outside of their activity group so that most students did not become acquainted with, if even introduced to, the other individuals from their permanent campus residence. Thus, at the time of the first data collection participants had known one another for only a few days. Although not a true zero-acquaintance paradigm, this design greatly reduced the amount of acquaintance between participants as compared to previous studies.

By the time of the final data collection period, participants had lived together for approximately 6 months. Most participants lived in fraternity houses while other groups were made up of roommates and nearby residents. The living arrangements of these students ensured a great deal of interaction during this time as they completed many tasks together, such as doing house chores, eating at the dining hall, participating in group activities (such as building Homecoming floats), and attending weekend retreats with other first-year students. Additionally, many of the freshmen shared common sleeping quarters which allowed for even greater interaction.

All data collection was conducted at the students’ residences. The researchers arranged to meet with all of the participants in each residence during a specific time, so that the vast majority of data collections took place with all group members present. Questionnaires were left with group members for those participants who could not meet at the scheduled time, and these students completed the questionnaires and returned them to the researchers via campus mail.

The multivariate round-robin design was analyzed using SOREMO (Kenny, 1993), a FORTRAN computer program created specifically for person perception analyses. This program conducts an SRM analysis that allows variance in people’s perceptions of others to be attributed to the three components of any person perception: a perceiver, a target, and the unique relationship between specific perceiver and target. The perceiver variance denotes the way in which an individual tends to perceive others. The target variance refers to the way in which an individual tends to be perceived. The relationship variance indicates the unique way in which an individual is perceived while at the same time controlling for the perceiver effects of the judge and the target effect of the individual. These variance terms are theoretical values estimating the variances that would be attributed to each source based on an infinite number of perceivers and targets. This theoretical value is the absolute variance estimate for each respective source. The relative variance estimate is a particular variance score, such as target variance, divided by the sum of the perceiver, target, and relationship variances (Because the absolute variances are theoretical values rather than estimates based on actual scores, the sum of the variance identified in perceptions is not restricted to a scale between 0 and 1. Relative variances, however, are restricted so that the identified variances sum to 1) Relative variances can be easily understood as correlations; for instance, the relative target variance is essentially the predicted correlation between the perceptions of two judges who rate the same targets. Consensus scores refer to the level of relative target variance in a set of judgments, thus indicating the level of agreement between individuals.

Since two indicators were used to measure each construct, error variance could be separated from relationship variance in these analyses. Thus, the sum of the variances is divided into four components, although only three are theoretically interesting. In SOREMO, variance is computed within each group and then pooled across groups. Single sample t tests were used to determine the significance level of each variance component. Significance tests use group as the unit of analysis for the variance components. The variance component used in each significance test was weighted by the group size less one.

As shown in the top panel of Table 1, participants were able to reach consensus with regard to specific trait ratings at each of the three time points. At Time 1, consensus of judgments was indicated with regard to extroversion, agreeableness, conscientiousness, and culture. The traits of emotional stability, creativity, and trustworthiness failed to reach statistical significance. At both of the later two time points, statistically significant levels of consensus were reached for each of the 7 personality traits rated. Overall, ratings of personality traits (judgments of the 7 traits each measured at 3 time points) yielded significant levels of interjudge agreement 18 out of 21 times. These findings show that individuals demonstrated a degree of consensus in judgments necessary to meaningfully examine the role that acquaintance may have in these agreement levels.

To assess changes in consensus as a function of increased acquaintance, a 7 (personality traits) x 3 (time of year) doubly multivariate repeated-measures MANOVA was conducted. This analysis revealed that across the time points of the study, the average levels of consensus differed with regard to some of the traits (see Figure 1). Specifically, statistical data on consensus differed significantly over time for judgments of conscientiousness (F(2, 56) = 4.12, p < .05), emotional stability (F(2, 56) = 3.06, p < .05), and differed marginally significantly in judgments of extroversion (F(2, 56) = 2.58, p < .10). To directly test the hypothesis concerning the effect of acquaintance on interjudge agreement, a priori contrasts were conducted to test changes in average consensus level between successive data collection periods. As illustrated in Table 2, consensus increased between Time 1 and Time 2 with regard to judgments of conscientiousness, t(28) = 5.75, p < .05, and emotional stability, t(28) = 4.35, p < .05. Interestingly, interjudge agreement was shown to decrease for perceptions of creativity between Time 2 and Time 3, t(28) = 4.00, p < .05. Judgments of no other traits demonstrated any changes in consensus across these time periods.

The present study was designed to examine whether interjudge agreement of personality traits increases as people become more acquainted. Whereas common beliefs and lay wisdom arguing that this relationship exists within naturally occurring settings have been backed by suggestive findings from research (e.g., Kenny et al., 1994), recent evidence gathered from more methodologically sound studies (e.g., Park et al., 1997) has been to the contrary. Specifically, a study with groups of college roommates who interacted in naturalistic settings over the course of eight months revealed that consensus levels in personality judgments did not increase with increased acquaintance (Park et al.). In the current study we built on the strengths of this research but also examined the initial period of acquaintance which had not been previously assessed. This examination provided evidence consistent with previous findings but also showed that agreement in personality judgments did change over the first weeks of acquaintance.

In accordance with common usage and understanding of the term, acquaintance was defined by marked periods of time. This chronological definition was validated by assessing participants’ subjective feelings of acquaintance at each data collection by directly inquiring how well they knew each group member. In this study, data collection times were determined to coincide with significant events within the naturalistic setting that allowed for varied interaction between group members (i.e., completing Homecoming events and finishing semester examinations). Not surprisingly, individuals reported knowing one another to a greater degree at each data collection; thus, measures of agreement at each data collection reflected an increase in acquaintance. Agreement was defined as consensus in the judgments of others and assessed using a SRM analysis (Kenny, 1993). The pattern of consensus reported at each data collection proved to be consistent with past person perception research. As illustrated in Table 1, there was a significant amount of target variance in people’s perceptions of one another. This significant level of target variance suggested that judges were able to reach a consensus regarding the specific standing of others on personality traits at each time and as early as during the first week of acquaintance

In the present study, analyses revealed that the abilities of individuals to agree in their judgments of the personality traits of others varied as a function of trait and time (see Table 2 and Figure 1). Specifically, agreement increased for judgments of conscientiousness and emotional stability between the first two time points. Consensus levels for the other five traits did not increase between Time 1 and Time 2. The finding that acquaintance affected agreement for some traits but not for others appears to be consistent with commonly held beliefs regarding the relationship between these variables. The belief that people understand one another better over a period of time implies that judgments concerning the personalities of others depend upon interactions with, or observation of, individuals in certain situations that highlight aspects of the personalities of those individuals. That is, to determine how reliable an employee is, a supervisor must first witness a situation in which that person demonstrates (or fails to demonstrate) behavior indicative of their reliability. This is consistent with the present finding that judgments of one’s emotional stability (defined as even-tempered or temperamental behavior), necessitated observations of the behaviors of people during various situations that demonstrated the consistency of their emotionality. Similarly, judgments of another person’s conscientiousness (as defined by responsible and reliable) required that opportunities to act in a responsible, or irresponsible, way were presented and observed. On the other hand, traits such as extroversion (outgoing and sociable) and culture (intelligent and knowledgeable) were apparent to judges in many interactions; therefore, judges made decisions about the standing of the individual along these dimensions more quickly and did not show the same increases seen in more situation-specific traits. The empirical findings can thus be seen to support the concepts that lay people have of the relationship between acquaintance and agreement.

However, consensus was not shown to increase with regard to any of the traits between the later two data collections. Failing to find increases in consensus levels after eight weeks of acquaintance was consistent with previous research. In their respective studies, Albright (as cited in Kenny et al., 1994) and Park et al. (1997) also failed to find, after similar periods of acquaintance, that increased acquaintance leads to increased agreement. Unexpectedly, the present data showed less agreement in judgments of creativity after 22 weeks than after eight weeks. Although neither predicted nor assessed in previous studies, this finding can also be interpreted as the opposite effect of the same dynamics that led to the earlier increases in agreement for emotional stability and conscientiousness. Because standards of creativity vary substantially there may be little basis to expect agreement in creativity judgments. Thus, increased opportunity to observe the creative works and thoughts of group members (measured as imaginative and creative) after the first eight weeks of acquaintance may have resulted in the use of different standards by judges rather than the baseline ratings that were used when less information was available earlier in the relationship.

These data conform perfectly with what was known from previous studies, but add to the previous understanding of the acquaintance process by extending the scope of the investigation. In this study we demonstrated that agreement level increased for specific traits between initial acquaintance and eight weeks later, but did not increase after eight weeks. Park et al. (1997) did not find an increase in consensus beyond two weeks of acquaintance. The discrepancy between the previous and present findings is, therefore, in the time period unique to the present study. Furthermore, this novel finding offers a reconciliation between empirical investigations and common knowledge. Given these findings, agreement does increase with acquaintance – at least initially. In fact, this finding suggests that greater change in agreement levels might be detected if zero acquaintance were to be operationalized even better than it was in the present study. Further research focusing more closely on judgments of traits within these first few weeks would provide an even better understanding of this critical acquaintance period.

While replicating previous findings indicating no effect of acquaintance, the current conclusions also provide evidence supportive of the popular belief that acquaintance does change agreement. Thus, extending the scope of study to include the earlier period of acquaintance allows for greater understanding of the entire process and helps clarify discrepancies between empirical research and common knowledge. These data show what people have known for years: agreement in personality judgments does increase with acquaintance, if only very early in the acquaintance process.